This invention relates to local area communication networks, and more particularly to local area networks (LAN) according to ANSI Fiber Distributed Data Interface (FDDI) standards, which LAN are arranged for enhanced reliability in adverse environments.
Modern vehicles such as aircraft, ships and the like include sophisticated computer systems for sensing the environment, evaluating the sensed information, and for controlling the vehicle and/or its payload in response to the evaluations. Such vehicles, and military equipment whether fixed or mobile, require highly reliable communications among the sensors, evaluation equipment, and the controlled devices. The reliability is often provided, at least in part, by redundancy. Thus, in addition to redundant computers for performing the evaluations, and redundant sensors and controlled devices, if possible, the interconnections should also be redundant. In warships, for example, there may be a relatively large number of sensors, including a plurality of radar, sonar, infrared and other sensors, all of which must be interconnected with command centers and with evaluation computers, and with guns, missiles and their controllers.
In the past, such warships have provided redundant point-to-point communication or data transmission paths or links between each mission-critical pair of devices, with the redundant paths differently routed within the ship to avoid the possibility that single-point damage can disrupt the communication. With this arrangement, all equipments remain interconnected, and any one device or sensor only becomes isolated from the system in the event of breaks on both of its separately routed, mutually alternate paths. Thus, extensive damage to the ship is likely to break several data paths, but these data paths are not likely to include both data paths to a particular device, so all devices are likely to remain interconnected. With the advent of computerized control, each computer (including the redundant computer) is required to provide a sufficient number of input-output (IO) ports to allow connection by at least first (primary) and second (alternate) paths to each of the devices which it serves. Specialized military computers with large numbers of input-output ports are available for such interconnections. The rapid expansion of control requirements tends to quickly render obsolete any computer with a fixed number of IO ports, and specialized computers with yet larger numbers of IO ports then become necessary.
With a view toward improved reliability and reduced cost, attention has been directed toward the use of a standardized local area network (LAN) in conjunction with a plurality of microprocessors, such as are in common commercial use, for performing the same functions as the special-purpose computers together with their point-to point communication paths. Such a LAN arrangement allows the distributed microprocessors to perform multiple functions, thereby providing another level of, and increasing the redundancy of the computing portion of the system, and also allows ready expansion of the system to include sensors or controlled devices as their need is discovered. The use of a standardized or commercial-type LAN further reduces cost, without, it is hoped, sacrificing usability.
The American National Standards Institute (ANSI) has generated standards and protocols for token-passing fiber-optic LAN systems, which are known by the acronym FDDI, which stands for Fiber Distributed Data Interface. It would be desirable, if possible, to use the FDDI standards in relation to high-reliability systems such as those described above. At this time, FDDI-based systems are being introduced into shipboard applications, as described, for example, in an article entitled "All Aboard FDDI", by Adams, published at pp 13-17 of the December, 1991 issue of Military and Aerospace Electronics magazine. However, the FDDI standards provide only limited redundancy, and so cannot be used for mission-critical communications.
In FIG. 1, a standard FDDI local area network includes first, second and third standard FDDI stations or connection nodes 1, 2 and 3. Each standard station includes first input and output ports 1i and 1o, respectively, and second input and output ports 2i and 2o, respectively. Input-output port 1i, 2o is also known as an A port, and input-output port 2i, 1o is also known as a B port. As illustrated in FIG. 1, the first output ports lo of each of the three standard stations 1, 2 and 3 are connected by fiber-optic transmission paths to the first input port 1i of the next standard station. For example, a transmission path 11 connects from output port 1o of station 1 to input port 1i of station 2. Similarly, a transmission path 12 connects from output port 1o of station 2 to input port 1i of station 3, and a transmission path 13 connects from output port 1o of station 3 to input port 1i of station 1, thereby forming a first loop, designated generally as 20, by which data can flow in the direction of arrow 24 among stations 1, 2 and 3. As also illustrated in FIG. 1, the second output ports 2o of each of the three standard stations 1, 2 and 3 are connected by fiber-optic transmission paths to the second input port 2i of the next standard station. For example, a transmission path 14 connects from output port 2o of station 1 to input port 2i of station 3. Similarly, a transmission path 15 connects from output port 2o of station 3 to input port 2i of station 2, and a transmission path 16 connects from output port 1o of station 2 to input port 2i of station 1, thereby forming a second loop designated generally as 22 by which data can flow in the direction arrow 26 among stations 1, 2 and 3. In this context, a transmission path is a path by which signal travels essentially unchanged from one location to another. A path may include cable, connectors, repeaters, switches, attenuators, and the like, the presence of which does not affect the signal.
Each station includes a Media Access Control unit which is associated with one, but not both, input-output ports, so that one of the two loops is designated as the main or primary loop.
The system as illustrated in FIG. 1 includes only three stations for simplicity of explanation, but could include a large number of such stations, connected in a similar manner. As illustrated, each station is connected to the next station by means of two fiber-optic transmission paths, as for example station 1 is connected to station 2 by transmission paths 11 and 16, in which data flows in mutually opposite directions. An envelope designated 28 represents the practical fact that the two fiber optic transmission paths 11 and 16 are often combined into a single multifiber cable. Similarly, envelopes 30 and 32 represent the combining of fibers 12 and 15 into a cable 30, and the combining of fibers 13 and 14 into cable 32. The FDDI network as so far described in conjunction with FIG. 1 is used by connecting sensors, computers (and their associated operator inputs) or controlled devices to externally accessible data paths associated with each station, such as path 58 associated with station 1, and paths 59 and 60 associated with stations 2 and 3, respectively. In operation, a token is produced at the primary output port of a station when it has transmitted all its current information, and the next station on the primary loop either seizes the token to make time available for its own transmission, or retransmits the token on the primary loop or ring to the next station when it has no data to transmit. Thus, the ability to transmit unimpeded on the common bus steps from station to station with the token. Reception is not dependent on the token, and all stations may, essentially simultaneously, receive data which is relevant to the stations.
On a vehicle such as a ship, it may be advantageous to route cable continuously from one end of the vehicle to the other end, connecting to various equipments along the way. This corresponds to that portion of FIG. 1 including stations 1, 2 and 3, and data paths included in cables 28 and 30, but without the data paths corresponding to those of cable 32 of FIG. 1. As described below, the two end stations "wrap" so that communications occur back and forth among the stations. This linear arrangement is termed a "root" topology The root topology has the advantage of reduced cabling requirement, but introduces the distinct disadvantage that a single break anywhere in the system results in undesirable isolation of portions of the system. This disadvantage could be remedied by a complete additional set of station nodes and interconnection cabling. However, this more expensive and complex than simply running the additional cable for the return link corresponding to cable 32 of FIG. 1.
Concentrators are allowed within the FDDI standard for expanded system connectivity. As described in an article entitled "Using Redundancy In FDDI Networks", by Ocheltree, published at pages 261-267 of the proceedings of the 15th Conference on Local Computer Networks, Minneapolis, Minn. Sep. 30-Oct. 3, 1990, and in an article entitled "The DECconcentrator 500 Product", by Tiffany et al., published at pp 64-75 of the Spring, 1991 issue of Digital Technical Journal, concentrators include a plurality of master ports in addition to the A and B ports. FIG. 2 illustrates a concentrator station in accordance with FDDI standards. In FIG. 2, an FDDI concentrator station 4 includes an A port with 1i and 1o ports, a B port including 2i and 2o ports, and a path 61 for connection to a local sensor or other device, just as in the standard FDDI stations of FIG. 1. In addition, concentrator station 4 of FIG. 2 includes a plurality of "master" (M) ports, which are available for connection of additional single stations 62a, 62b, 62c, . . . 62n. Such an arrangement might have utility, for example, in a LAN in which a multi-story building is to be interconnected. For this purpose, each floor might have a concentrator station, with the A and B ports connected with other concentrators on other floors in a loop as described in conjunction with FIG. 1, and at each floor, the M ports of the concentrator for that floor are connected by individual two-fiber cables to each separate single station For example, if concentrator 4 of FIG. 2 were on the fourth floor of such a building, port M1 would be connected to fourth-floor equipment station 62a by way of a cable 63a, port M2 would be connected to fourth-floor equipment station 62b by way of a cable 63b, etc. The A and B ports of the concentrator would, of course, be connected to other stations on the third and fifth floors. In the terminology of FDDI, the additional single stations 62 of FIG. 4 are part of a "tree", which in the case of single stations as in FIG. 4 cannot be expanded to further levels. If, in FIG. 2, each single station 62 were to be replaced by a concentrator station which was connected to its higher-level concentrator station (station 4 in FIG. 1) at one of its A or B ports, additional stations at a third level could be connected to the various M ports of the second-level station, to thereby create a tree with many levels. In such an arrangement, the lower-level stations all couple back to loop through concentrator 4, and are termed "singly homed". On the other hand, if the other of the A and B ports of the second-level concentrator is connected to another concentrator station which is also in the loop, each low-level station can couple to the loop through two (or possibly more) paths, and they are said to be "dual homed", all as described, for example, in the above mentioned Ocheltree article. In operation of the concentrator as illustrated in FIG. 2 when connected in a ring, as described, the token is received at the A1i port as in FIG. 1, and is routed in succession to each of the active M ports and to the stations connected thereto, i.e. from port B to port M1, then down through all the stations in the M1 path, then to M2, and through all the stations in the M2 string,... When the token has progressed through all the stations associated with all the M ports of the concentrator, it is made available at the B1o output port on the ring. Thus, the use of concentrators in a loop gives rise to delays which do not occur in standard stations, and their use is contraindicated unless necessary It should be noted that the paths between M ports of a concentrator station and the next following station are under continuing self-test during those intervals in which it is not being used for data.
FIG. 3a illustrates one possible configuration of bus connections within a standard or concentrator station, taken as standard station 1 for definiteness. In FIG. 3a, first and second directional couplers 40 and 42, respectively, have their main through lines coupled to each other and to input port 1i and output port 1o. The branch lines 44 and 46 of directional couplers 40 and 42, respectively, are coupled to active circuits illustrated as a block 56, which may include light detectors and modulators. Similarly, third and fourth directional couplers 50 and 48 have their through paths coupled to each other and to input and output ports 2i and 2o, respectively, and their coupled branch lines coupled to active circuit block 56. Suitable directional couplers include the well-known star couplers. Those skilled in the art know that this arrangement couples signals directly from input port 1i to output port 1o, and from input port 2i directly to output port 2o, and also couples the active circuit portion 56 to receive signals from the input ports 1i and 2i of standard station 1, and to couple signals therefrom to output ports 1o and 2o of station 1, but has the disadvantage of splitting the transmission power and thereby reducing the signal-to-noise ratio. In many applications, this will be of no consequence, but when large numbers of stations are to be interconnected, the cumulative loss of the couplers may exceed the regenerative power of the various stations.
Lower through loss at each station is permitted by the use of a switched "Trunk Coupling Unit" (TCU), illustrated in FIG. 3b. In FIG. 3b, station 1 receives signal through input port 1i, which is applied to a moving-mirror fiber-optic switch 98. Switch 98 allows signal to be routed directly from input port 1i to output port 1o when that station or node is shut down or deprived of power, which is a fail-safe bypass mode of operation. In normal operation, when the station is powered up, switch 98 couples signal from input port 1i to the active portions 56 of the node, and from the active portion to output port 1o. The operation of the TCU associated with ports 2i and 2o is similar.
The FDDI standard provides for error recognition, and for reconfiguring the station in response to errors for improving reliability. FIG. 4a is similar to FIG. 1, but includes an assumed break in a transmission line. While standard stations are described, the following discussion is equally applicable to concentrator stations. As mentioned, the transmission lines are actually coupled into cables including at least two paths, as for example cable 28 includes the two transmission paths 11 and 16 extending between stations 1 and 2. As a result, a break in one cable almost invariably breaks both optical fibers therein. The FDDI standard takes cognizance of this likelihood, and arranges its error detection circuits within each station to recognize the existence of faults, which include (a) a fault in the transmission path terminating at a 1i input port, and (b) a fault in the transmission path which starts at a 1o output port. The (a) type of fault may be termed an "upstream" fault, and the (b) fault may be termed "downstream", relative to the "main" data path or loop. In FIG. 4, a large "X" is placed over multifiber cable 28, thereby indicating that it is broken, possibly as a result of environmental or hostile action. As to station 1, the break is a downstream "b" fault, and as to station 2, it is an "a" type upstream fault. The FDDI standard requires the station subject to an upstream fault (in this example, station 2) to disconnect its 2i input from its 2o output and reconnect or "wrap" it to its 1o output, as illustrated by dashed path 70 in station 2. The disconnection of input port 2i from output port 2o is of no consequence, because any data applied to output port 2o cannot traverse data path 16 of cable 28 anyway. The FDDI standard also prescribes that a station subject to a downstream fault (station 1 in the example) must disconnect its 1i input port from the 1o output port, and reconnect or wrap it to the 2o output port, as illustrated by dash line 72 in station 1. As in the case of station 2, disconnecting input port 1i from output port 1o makes no difference, as data applied to data path 11 of cable 28 cannot traverse the break. The FDDI wrap protocol results in formation of a third loop to replace the broken first loop 20 and second loop 22. The third loop extends from output port 1o of station 2, through path 12, ports 1i and 1o of station 3 to path 13, then through input port 1i, path 72, and output port 2o of station 1, through path 14 to input port 2i of station 3, and out of output port 2o of station 3, through data path 15 to input port 2i of station 2, and finally through path 70 back to output port 1o of station 2. Thus, a single break has no significant effect on the communications among stations 1, 2 and 3 (although the delay may increase somewhat).
FIG. 4b is similar to FIG. 4a, differing only in that a second break in the cabling is shown, illustrated by a large "X" over cable 30. This break obviously isolates station 2 from stations 1 and 3. Station 1 still experiences a downstream fault, and responds by producing internal reconfiguration path 72. Station 3 experiences an upstream fault, and responds by reconfiguring itself by wrapping its 2i input port to its 1o output port, just as station 2 did in FIG. 4a. The wrap path in station 3 is designated 74. The wrappings provided by reconfiguration paths 72 and 74 of stations 1 and 3, respectively, form yet another loop which includes stations 1 and 3, and paths 13 and 14. In addition to the isolation of station 2 from stations 1 and 3 by the breaks of cables 28 and 30, station 2 is further isolated by the operation of the wrap specified by the FDDI protocol, which requires disconnection of paths 11 and 15 from station 2. As mentioned, the illustrated system includes only three stations, for simplicity of explanation. It should be understood that station 2 of FIG. 4b may actually represent a plurality of stations, and that the wrap specification will result in those additional stations forming their own interconnected loop, which, however, is totally isolated (by the multiple breaks) from the loop including stations 1 and 3. For mission-critical data communications, isolation of whole sections by damage to two unrelated data paths may not be acceptable. It should be noted that the looping data paths in an FDDI system are often illustrated side-by-side, as is shown by the alternate placement of data path 14 in the form of a dot-dash line 14b adjacent to data path 13 in FIG. 4b. This configuration results in the use of the terminology of formation of a "U" or "horseshoe" path instead of an alternative "loop" path when a wrap occurs. Thus, reconfiguration forms a horseshoe or a loop, depending upon the form or topology of the circuit which one has in mind, but the horseshoe and the loop are equivalent terms in this context.
Thus, while it would be desirable to use the FDDI system for all shipboard communications, including mission-critical data, the possibility of total isolation of portions of the communications system due to multiple breaks in the transmission cable makes it unusable for mission-critical applications. It might be possible to use an FDDI system for auxiliary communications, while using point-to-point communications for mission-critical applications, but this would not help the key problems of expandability of an existing system, cost, and the need for special-purpose computers. If the FDDI system standards are modified to achieve the reliability goal, the cost benefits of commonality with commercial systems are lost.
A SAFENET study group established by the Navy has conducted a review of LAN for mission-critical applications, and is in the process of generating SAFENET I standards based on IEEE 802.5 "free-token" protocol, which standard specifies that the two fiber transmission lines be separately routed, and adopts "ring hop", which amounts to abandoning a defective loop or ring and using the operational loop. The free-token protocol is not considered advantageous, because the hold time at any station is not controlled, and thus priority messages may be delayed excessively. The SAFENET group has also at least partially generated SAFENET II LAN standards. The SAFENET II standards define a timed token ring pursuant to FDDI standards, but in which the dual counter-rotating rings are separately routed. It supports a single failure, but degrades in the presence of additional failures, and therefore is no better than FDDI.
An improved communication system is desired.